Error while Finetuning the Trained Model

[almakhmudov]

Dear NequIP community,

Recently I've been struggling with the finetuning of the previously trained model on another dataset. I've included the "initialize_from_state" as well as the "initial_model_state" that points to best_model.pth in the config file. However, I keep getting the following error:
Traceback (most recent call last):
File "/home/albert/.conda/envs/nequip/bin/nequip-train", line 8, in
sys.exit(main())
File "/home/albert/.conda/envs/nequip/lib/python3.8/site-packages/nequip/scripts/train.py", line 94, in main
trainer = fresh_start(config)
File "/home/albert/.conda/envs/nequip/lib/python3.8/site-packages/nequip/scripts/train.py", line 222, in fresh_start
final_model = model_from_config(
File "/home/albert/.conda/envs/nequip/lib/python3.8/site-packages/nequip/model/_build.py", line 179, in model_from_config
model = builder(**params)
File "/home/albert/.conda/envs/nequip/lib/python3.8/site-packages/nequip/model/_weight_init.py", line 29, in initialize_from_state
return load_model_state(
File "/home/albert/.conda/envs/nequip/lib/python3.8/site-packages/nequip/model/_weight_init.py", line 62, in load_model_state
graph_model.load_state_dict(state, strict=strict)
File "/home/albert/.conda/envs/nequip/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1497, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for GraphModel:
Unexpected key(s) in state_dict: "model.model._empty".

Any help would be very much appreciated!

Best,

Albert

[cw-tan]

You should be able to do this easily with new v0.7.0 update, which will be very different from the old infrastructure. Anyway, this section of the docs is relevant for fine-tuning.